I use it at (SUSE Linux Enterprise Server, at work); I never particularly liked or disliked it; but once they adopted systemd - I lost respect for it.
Either way, doesn't every major Linux distro other than Gentoo use systemd now?
(There is also Devuan, which is "Debian without systemd", i.e. minimal changes to allow for systemd not to be used, but of course that doesn't change the default for everything depending directly on Debian.)
There are all sorts of complaints against systemd, some may or may not be overblown - I certainly have not investigated many use cases - but the problem is with the core of it: A single software project gradually taking over more and more of the OS userland. That is simply not a good thing to stick on your system.
No less importantly, I don't believe systemd solves any significant problems vis-a-vis its alternatives.
About "making Linux better" - that's false on principle, on architectural grounds, and is not a matter of finding.
- Timer units prevent long job duplication and have mechanisms for restarting as soon as possible if an execution is missed. Cron can't do that on its own, and scripting such functionality is brittle. - Service units are much simpler and easier to understand than shell script init, restart handing is very good to have built in, and dependency management is extraordinarily useful. User units are also great. - Automatic capturing of logs with journald and its built-in disk usage limitations are better in practice than piping to files/syslog and running logrotate on blind schedules. - cgroup resource controls for processes are far more effective than nice/ionice, etc, and can be altered on the fly for every unit without restarting. - systemd-resolved is one of the few ways to get system-wide support for dns-over-tls, dnssec, and per-service/network choice of DNS servers without considerable effort.
These are just the features that I personally see and interact with almost daily, but I have seen and used enough to know that its dbus and udev integrations are stellar and enable useful functionality for desktops and networking.
I understand that some people have decided that religious adherence to a vague unix dogma is more important than features, usability, and reliability, but please accept that some of us have work to do, which includes the maintainers of major Linux distributions.
Perhaps because these people consider that solving all the problems that systemd claims to solve does introduce too many other problems, so they tend to consider it to be the better solution to solve, say, 90 % of the problems that systemd solves, but introduce less new problems.
My recent experience was trying out Fedora atomic. I love that idea. I found systemd is kinda nice for service management.
But still, I kept running into issues with it spreading into everywhere, doing its own thing, and difficulty working around it. Partially figuring out the atomicity, and partially that no other distro I've tried has leaned in that hard to systemd.
I'm moving on to Arch since it looks like I can at least get this out of my boot process.
(Or, more on the topic of the thread, Tumbleweed looks like an interesting take on being able to roll back to known working states as a replacement for Fedora atomic.)
There's also SixOS coming soon (https://events.ccc.de/congress/2024/hub/de/event/sixos-a-nix...). NixOS does kind of a similar atomic thing, so you might enjoy that, but vanilla NixOS is systemd based, so once SixOS drops you might get exactly what you want.
I saw some of that SixOS and I am really interested. There was an "ownerboot" tool they linked to that also looked nice to me.